How Sigma Delta Modulators achieve

high performance
(and why they arent even better)
Josh Reiss
josh.reiss@elec.qmul.ac.uk
Centre for Digital Music
Queen Mary, University of London
UK

signal-to-noise ratio (SNR)

single most used performance characteristic in A/D converters

ratio of rms (root mean squared) signal to rms noise within the
bandwidth of interest
20log10 of this ratio to derive SNR in decibels
Or 10log10 of signal power over noise power

SNR can be found for any A/D converter

Approach
explain the basics of the signal-to-noise ratio (SNR),
theory and estimation in PCM and SDM converters
derive the formula for the SNR of an arbitrary SDM

Show simulated & theoretical SNR as a function of bits, order, OSR,

input amplitude
Show how a high SNR is obtained
Show what prevents SDM implementations from achieving theoretical values

Signal-to-noise ratio is probably the single most used performance characteristic in

A/D converters. Here, we will explain the basics of the signal-to-noise ratio (SNR),
its theory and estimation in PCM and SDM converters, and how it can be estimated
in various situations.
The SNR is given by the ratio of rms (root mean squared) signal to rms noise within
the bandwidth of interest. You then multiply the log10 of this ratio by 20 to derive
SNR in decibels. The SNR can be found for any A/D converter. First, we derive the
formula for the SNR of an ideal analog-to-digital converter.

Quantization Assumption

quantization levels - allowed values

in output signal after quantization
quantization step size q- distance
between 2 successive levels
quantiser with b bits covering range
from +V to V
2b quantization levels
width of each quantization step is

q=2V/(2b-1)
quantization error - difference
between input & output to quantiser
eq=Q(x)-x
rounding quantizer - assigns each
input to nearest quantization level

1.0
0.8
0.6
0.4

Q(x)

2/(2^3-1)=2/7

0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

Transfer curve for b=3, V=1

q / 2 eq (n) q / 2

The allowed values in the output signal, after quantization, are called quantization
levels, whereas the distance between 2 successive levels, q, is called the
quantization step size. Consider a quantiser with b bits covering the range 2V, from
+V to -V. Then there are 2b quantization levels, and the width of each quantization
step is
q=2V/(2b-1)
This is depicted for a 3 bit quantiser and V=1.
The rounding quantizer assigns each input sample x(n) to the nearest quantization
level. The quantization error is simply the difference between the input and output to
the quantiser, eq=Q(x)-x. It can easily be seen that the quantization error eq(n) is
always bounded by +/- q/2

Quantization Assumption (2)

q / 2 eq (n) q / 2

Assume quantisation error uniformly

distributed over a quantisation step

1/ q | eq | q / 2
p (eq ) =
0 | eq |> q / 2

p(eq)
1/q

not exact

q=2V/(2b-1)

still quite close to accurate for noisy

and some sinusoidal signals

-q/2

+q/2

eq

Probability distribution function

for the quantization error.

We assume that the quantisation error is uniformly distributed over a quantisation

step. This assumption is not exact, but is still quite close to accurate for sinusoidal
signals. The pdf, probability distribution function for the quantization error, is
depicted.

SNR for a PCM A/D Converter

If sampling rate satisfies sampling theorem, fs>2fB, quantization is only error

in A/D conversion process
jitter and other effects are not considered here

Assuming uniform distribution, average quantization noise is given by

eq = E{eq } =

q/2

eq p (eq )deq =

quantization noise power given by

= E{(eq eq ) } = E{eq } =
2
e

q / 2

Or

e2 =

eq

1
deq = 0
q
q/2

eq p (eq ) deq =
2

eq 2

q / 2

1
q2
deq =
q
12

q 2 (2V /(2b -1)) 2

V2
V2
=
=

12
12
3(2b -1) 2 3 22b

estimate signal power

assume we quantize sinusoidal signal of amplitude A, x(t)=Acos(2t/T)
T

x2 = E{(ex ex ) 2 } = E{ex 2 } =

1
A2
2
(
cos(2
/
))
A

t
T
dt
=
2
T 0

If the sampling rate satisfies the sampling theorem, fs>2fB, quantization is the only
error in the A/D conversion process (jitter and other effects are not considered
here). Using the assumption of uniform distribution, we can find the average
quantization noise and the quantization noise power.
Quantisation noise power is found using textbook definitions
To find the SNR, we also need to estimate the signal power. Now assume we are
quantizing a sinusoidal signal of amplitude A, x(t)=Acos(2pt/T). The average power
of the signal is thus A2/2
From and , the signal-to-noise ratio may now be given by,
Thus the signal-to-noise ratio increases by approximately 6dB for every bit in the
quantiser. Using this formula, an audio signal encoded onto CD (a 16 bit format)
using PCM, has a maximum SNR of 98.08. Also note from , that the SNR is linearly
related to the signal strength in decibels.

SNR for a PCM A/D Converter (2)

Quantization noise power

Signal power

signal-to-noise ratio,

SNR(dB)=20log10

e2 =

V
V

3(2b -1) 2 3 2 2b

x2 = A2 / 2

x
A
20log10 + 6.02b + 1.76
e
V

increases by 6dB for every bit in quantiser

linearly related to signal strength in decibels

Audio signal encoded onto CD (16 bit format) using PCM

maximum SNR of 98.08.

The signal-to-noise ratio may now be given.

Thus the signal-to-noise ratio increases by approximately 6dB for every bit in the
quantiser. Using this formula, an audio signal encoded onto CD (a 16 bit format)
using PCM, has a maximum SNR of 98.08. Also note from , that the SNR is linearly
related to the signal strength in decibels.

Oversampling
Previously
highest

assumed
input frequency was fB, signal acquired at Nyquist rate, fs=2fB.

assume signal oversampled, fs=2r+1fB

Oversampling - sampling input signal at frequency much greater than nyquist
now

decreases

quantization noise in band of interest

pe(f)
Nyquist rate PCM, f s/2=f B
4x Oversampled PCM, f s/2=4f B
8x Oversampled PCM, f s/2=8f B

fB

4fB

8fB

SNR for an Oversampled PCM A/D Converter

oversampling ratio is OSR=2r=fs/2fB.

quantization noise spread over larger frequency range
only concerned with noise below Nyquist frequency

in-band quantization noise power given by

q/2

2f 1
q2
n2 = eq 2 p (eq ) deq = eq 2 B deq = r
= e2 / OSR
f
q
2
12

q / 2
Most noise power is now located outside of signal band
signal power occurs over signal band only

SNR now given by

remains unchanged

SNR=10log10 ( x2 / n2 )=10log10 ( x2 / e2 ) + 10log10 2r

20log10 ( A / V ) + 6.02b + 3.01r + 1.76

For every doubling of oversampling ratio, SNR improves by 3dB

6dB improvement with each bit in quantiser remains
doubling oversampling ratio increases effective number of bits by a bit

The above discussion assumed that the highest possible frequency in the input
signal was some value fB and that the signal is acquired at the Nyquist rate, fs=2fB.
However, lets now assume that the signal is oversampled such that the rate is
fs=2r+1fB, that is, the oversampling ratio is OSR=2r=fs/2fB. Thus, the quantization
noise is spread over a larger frequency range yet we are only concerned with noise
below the Nyquist frequency.
The in-band quantization noise power is now given by the total quantisation noise
power divided by the OSR.
Most of the noise power is now located outside of the signal band. The signal power
occurs over the signal band only, so it remains unchanged.
The signal-to-noise ratio may now be given by, the previous SNR +10log10OSR
Thus for every doubling of the oversampling ratio, the SNR improves by 3dB. The
6dB improvement with each bit in the quantiser remains, so we can say that
doubling the oversampling ratio increases the effective number of bits by a bit.

SNR for an Oversampled PCM A/D Converter

In-band quantization noise power can also be given by

n2 =

fB

Se ( f ) | NTF ( f ) |2 df =

fB

e2
fs

fB

| NTF ( f ) |

df

fB

where power spectral density of (unshaped) quantization noise is

Se ( f ) = e2 / f s

no noise shaping, just oversampling

noise transfer function is uniformly one over range [-fB,fB]

=
2
n

e2
fs

fB

fB

1df =

2 e2 f B
= e2 / OSR
fs

Same as before.
Equivalent of 16 bit Nyquist rate PCM (98dB SNR)
8bit 2.64GHz

As a lead-in to the following sections which deal with sigma delta modulation,
consider that the (in-band) quantization noise power can also be given as a function
of the Noise Transfer Function and the power spectral density of the (unshaped)
quantization noise. Since there is no noise shaping, just oversampling, we have that
the noise transfer function is uniformly one over the range [-fB,fB]. So we get the
same as .before

Oversampling
Doubling the oversampling ratio results in
approximately a 3dB drop in the noise floor

12 bit PCM, full scale input at 11.025kHz, triangular PDF dither

10

SNR for a 1st Order Sigma Delta Modulator

pulse code modulators perform no noise shaping whatsoever

first order sigma delta modulator.
E(z)

X(n)+

U(n+1)
+

U(n)
Q

Delay

Q(n)

X(z) +
+
-

Time-domain block diagram

Y(z)

-1

z
1-z-1

z-transform block diagram, quantiser

approximated by noise source

U (n + 1) = X (n) Q(n) + U (n)

given as

Use linear model of SDM

Time domain equation now given as

Quantiser modeled as noise source, e=Q-U

Q(n) = X (n 1) + Q(n) U (n) Q(n 1) + U (n 1)

= X (n 1) + e(n) e(n 1)

11

SNR for a 1st Order Sigma Delta Modulator

first order sigma delta modulator

E(z)

X(n)+

U(n+1)
+

U(n)
Q

Delay

X(z) +

Q(n)

Time-domain block diagram

Y ( z ) = X ( z ) z 1 + E ( z )(1 z 1 )

signal transfer function z-1

noise transfer function 1-z-1

Y(z)
+

z-transform block diagram, quantiser

approximated by noise source

Q(n) = X (n 1) + e(n) e(n 1)

z-1
1-z-1

So,

NTF ( f ) = 1 e j 2 f / fs
| NTF ( f ) |2 = ... = 4sin 2 ( f / f s )

=
2
n

fB

fB

Se ( f ) | NTF ( f ) | df =
2

e2
fs

fB

4sin 2 ( f / f s )df

fB

Here, Se(f) is the power spectral density of the unshaped quantization noise

12

SNR for a 1st Order Sigma Delta Modulator

In-band, shaped quantization noise power

2
2

n2 = e2 2 / OSR sin( / OSR)

e2
3

3 OSR
SNR(dB)

10log10

x2
A
20log10 + 6.02b + 9.03r 3.41
2
V
n

9dB for each doubling of the oversampling ratio

Compare with 3dB improvement which occurs without noiseshaping

Equivalent of 16 bit Nyquist rate PCM (98dB SNR)

1bit 96.78MHz

The effect of first order noise shaping is evident. We now get an improvement of
9dB for each doubling of the oversampling ratio, rather than the 3dB improvement
which occurs without noiseshaping.

13

SNR for a 2nd Order Sigma Delta Modulator

E(z)
X(z) +

+
+

Y(z)
z-1

+
-

z-1

Transfer function

Y ( z ) = X ( z ) z 1 + E ( z )(1 z 1 ) 2 = STF ( z ) X ( z ) + NTF ( z ) E ( z )

Noise Transfer function

NTF ( f ) = (1 e j 2 f / fs ) 2 = [2sin( f / f s )]2

In-band, shaped quantization noise power

n2 =

e2
fs

fB

[2sin( f / f )] df
4

fB

4
5 OSR

e2

14

SNR for a 2nd Order Sigma Delta Modulator

n2

4
5 OSR

e2

SNR (dB)

x2
A
20log10 + 6.02b + 15.05r 11.14
V
n2
15dB improvement in the SNR with each doubling of the
oversampling ratio
Equivalent of 16 bit Nyquist rate PCM (98dB SNR)

10log10

1bit 6.12MHz

Compared with the 1st order SDM, this provides more suppression of the
quantization noise over the low frequencies, and more amplification of the noise
outside the signal band.
Thus we see a large improvement in moving to a second order SDM. There is now
a 15dB improvement in the SNR with each doubling of the oversampling ratio.

15

SNR for an Nth Order Sigma Delta Modulator

Transfer function of a generic Nth order SDM, is given by

Y ( z ) = X ( z ) z 1 + E ( z )(1 z 1 ) N = STF ( z ) X ( z ) + NTF ( z ) E ( z )

Noise power in the baseband is given by

NTF ( f ) = (1 e j 2 f / fs ) N = [2sin( f / f s )]N

8
Oversampled PCM
1st Order SDM
2nd Order SDM
3rd Order SDM
Cutoff Frequency

Noise Transfer functions

for 8 times OSR

Magnitude Spectrum

7
6
5
4
3
2
1
0
20

40

60

80
f (kHz)

100

120

140

160

The transfer function of a generic Nth order SDM, is given by

The noise power in the baseband is given by as plotted.

16

SNR for an Nth Order Sigma Delta Modulator

NTF ( f ) = (1 e j 2 f / fs ) N = [2sin( f / f s )]N

Using an integral formula, this gives

n2 =

Giving

fs

fB

[2sin( f / f s )]N df e2

fB

SNR(dB)=10log10
20log10

e2

2N
(2 N + 1)OSR 2 N +1

x2
(2 N + 1)2(2 N +1) r
+ 10log10
2
e
2N

A
+ 6.02b + 1.76 + 10log10 (2 N + 1) 9.94N + 3.01(2 N + 1)r
V

In general, for an Nth order SDM

3(2N+1)dB improvement in the SNR with each doubling of the oversampling
ratio
6dB improvement with each additional bit in the quantiser
use of high order SDMs & high oversampling ratio offers much better SNR
than simply increasing # bits

The transfer function of a generic Nth order SDM, is given by

The noise power in the baseband is given by
Using an integral formula, this givesGiving
Thus in general, for an Nth order SDM, there is a 3(2N+1)dB improvement in the
SNR with each doubling of the oversampling ratio, and a 3dB improvement with
each additional bit in the quantiser. Thus, use of high order SDMs and a high
oversampling ratio offers a much better SNR than simply increasing the number of
bits.
Of course, this is an approximation. It depends on the coefficients of the modulator,
on the approximations used in the derivation, and other factors. Nevertheless, It
provides an upper limit on the SNR, and many sigma delta modulators perform
quite close to this limit.
For a 64 times oversampled 1-bit A/D converter, using a fifth order SDM, such as is
typical in audio applications, we find that
I dont believe that anyone has ever designed a 5th order, 1 bit SDM which gives
such a high SNR for a 64 times OSR. This is mainly because a design with near
ideal noise shaping characteristics would be too unstable to allow any realistic input.
But they have achieved SNRs of 120dB.

17

SNR for an Nth Order Sigma Delta Modulator

this is an approximation
depends on
coefficients of the modulator
on approximations used in derivation
Assume high oversampling ratio
High number of bits in quantiser

other factors.

provides rough upper limit on SNR

Some low order sigma delta modulators perform quite close
to this limit

18

SNR vs Bits
120

Predicted from

20log10

A
+ 6.02b + 1.76
V

assumes
V2
V2
=

b
2
3(2 -1)
3 22b
2
e

110

x
e

100

Simulated SNR
Predicted SNR

90
SNR (dB)

SNR(dB)=20log10

80
70
60
50
40
30
20
10
2

10
12
# of Bits

14

16

18

20

SNR as function of # bits in quantiser for PCM encoded

signal, sampled at Nyquist frequency
input freq. 2kHz , full range -1 to 1, sampling rate 44.1kHz

19

Effect of bits and signal amplitude

5th Order, 64 times OSR

From the formula using

250
240
230

SNR

220

e2 =

V
3(2b -1) 2

210
200
190

Input Amplitude
0.1
0.6
0.2
0.7
0.3
0.8
0.4
0.9
0.5
1.0

180
170
160
150

6
7
8
9
Number of Bits

10

11

12

13

14

15

16

20

SNR vs Oversampling Ratio

Assumptions
1. high OSR
lack of agreement for 1, 2,
and 4 times oversampling.

200
SNR (dB)

2. Precision issues for high

SNR
disagreement between
theoretical & simulated
results for 2nd order SDM at
OSR=512 or 1024

PCM, Simulated SNR

PCM, Predicted SNR
1st order SDM, Simulated SNR
1st order SDM, Predicted SNR
2nd order SDM, Simulated SNR
2nd order SDM, Predicted SNR

220

180
160
140

3. uniform noise
distribution
constant error between
theory & simulation for PCM

120
100

3, 9 & 15dB increases for

doubling OSR confirmed

5
6
log2(OSR)

10

21

Why arent SDMs close to the ideal?

Example
64 times oversampled 1-bit A/D converter, using 5th order SDM

SNR(dB) 20log10

A
+ 167.15
V

No 5th order, 1 bit SDM gives such high SNR for 64 times OSR

Many assumptions, but almost all lead to only minor

differences
#bits>>0
OSR>>0
Uniform quantisation noise distribution

Design with near ideal noise shaping characteristics would

be too unstable to allow any realistic input
But they have achieved SNRs of 120dB.

22

Shameless promotion
AES 31st International Conference

New Directions in High

Resolution Audio
London, UK
June 25-27, 2007
www.aes.org/events/31

23